Message boards :
Number crunching :
HadCM3 Full Resolution model low credits
Message board moderation
Previous · 1 · 2
Author | Message |
---|---|
Send message Joined: 28 Oct 04 Posts: 64 Credit: 34,444,555 RAC: 0 |
I am looking at the reported stats from CPDN for my systems as sent to [http://climateapps2.oucs.ox.ac.uk/cpdnboinc/hosts_user.php?sort=name&rev=0&show_all=0&userid=27607] Six day list: date S0 S1 S2 S3 S4 S5 S6 S7 S8 24th 331 394 1541 1007 588 976 603 1573 S9 S10 S11 1632 1318 639 25th 272 324 1507 985 482 1004 506 1566 1656 1541 524 26th 246 293 1517 975 437 1025 489 1592 1711 1631 475 27th <no rec> 28th 202 240 1537 957 371 470 286 1120 1856 1653 389 1835 29th 183 218 1487 949 361 426 283 1130 1874 1702 353 1855 30th 166 197 1432 941 383 386 281 1150 1901 1665 319 1825 I can't see any consistent positive pattern except for S8. Maybe the credits are just not delivered yet, though I check almost every day to make a record. If you have a clue, please let me know. :-} BillN |
Send message Joined: 5 Sep 04 Posts: 7629 Credit: 24,240,330 RAC: 0 |
Assuming (never a good idea with limited data), that S0 is this computer (1133710), then it probably has little to do with "credit rates". That computer "appears" to be running mostly regional models, which, as you can see, have been given the Status of No Resubmission. This new status label has caused problems, as explained in my post in News and Announcements. It's likely that, in your case at least, that no new credits are being granted. Which probably would be case with everyone. And if this IS the problem with your drop in credits, then Tuesday morning is the earliest that an Official Permanent Fix can and will be made. To everyone's credits. Backups: Here |
Send message Joined: 28 Oct 04 Posts: 64 Credit: 34,444,555 RAC: 0 |
Thanks again for the response, your followup is appreciated. I have also seen variable results in what appear to be good working systems, but until Wednesday or so, I'll wait for results and hope the problems vanish silently in the night. :-} BillN |
Send message Joined: 28 Mar 11 Posts: 35 Credit: 82,588 RAC: 0 |
I don't know why this is happening, so I will take a look, and maybe come up with an 'Official Permanent Fix' (sounds ominous :-( ). |
Send message Joined: 12 Feb 08 Posts: 66 Credit: 4,877,652 RAC: 0 |
There is something wrong with the Recent Average Credit. I am currently only crunching HadCM3N models and I have been steadily generating about 2500 credits per day. However, according to CPDN, my RAC is only 1300. It dropped there from 2500 when I switched to CM3N models. http://boincstats.com/stats/user_graph.php?pr=cpdn&id=501494 |
Send message Joined: 12 Sep 04 Posts: 34 Credit: 1,017,702 RAC: 0 |
I don't know why this is happening, so I will take a look, and maybe come up with an 'Official Permanent Fix' (sounds ominous :-( ). That would be appreciated. These models are definitely yielding less credit. If adjusted, at least it will make us feel good about selecting these models and help meet some of the deadline requirements. |
Send message Joined: 3 Nov 10 Posts: 39 Credit: 2,494,427 RAC: 0 |
...indeed - a little "vigorish" of about a 10% bonus in the credit department would encourage folks to run the 50-day workunits... |
Send message Joined: 19 Apr 08 Posts: 179 Credit: 4,306,992 RAC: 0 |
Maybe this is akin to the RAC issue HADAM3P's had when they were first introduced. Although in that case it was too much credit. It was beaten to death in this thread. Milo never revealed his fix. |
Send message Joined: 19 Apr 08 Posts: 179 Credit: 4,306,992 RAC: 0 |
@Warped: I like the multiverse explanation, but the Loch Ness Monster one is wanting: I would like to know: (link to original post) edited for link typo |
Send message Joined: 1 Jan 07 Posts: 1029 Credit: 36,058,265 RAC: 22,560 |
Maybe this is akin to the RAC issue HADAM3P's had when they were first introduced. Although in that case it was too much credit. It was beaten to death in this thread. Milo never revealed his fix.He "divided trickle_timestep by 3.7506944444444446 for all models with id 9". There was a better, and more mathematically grounded, explanation that generated that correction factor, but it would have caused problems elsewhere in the database if applied 'raw'. @ Jonathan: I can point you to that reference if you need it. |
Send message Joined: 31 Aug 04 Posts: 391 Credit: 219,896,461 RAC: 649 |
No clue here. on cpdn have RAC=4700 at BOINCstats have 8,600 Somewhere between 8K and 9.5 K is my long term average and I haven't put a new machine online for over a year. Some newb at CPDN will figure the BOINC credit anomaly when they have time. My only tries to understand the BOINC code have been weak and pitiful. Keep crunching edit -- PS the souce code is out there at BOINC -- but me no grok it nohow |
Send message Joined: 28 Oct 04 Posts: 64 Credit: 34,444,555 RAC: 0 |
Well, here it is Friday and it appears that the scoring fix is not in yet, at least from my systems results. Of 12 systems, two are offline, working systems: three are up slightly, and *seven* are down in small to medium amounts. I did see the PHKUP machine was down for a day or two, so that may be part of the trouble. It's back up as I just checked. Looks like we have a little longer to wait. :-{ BillN |
Send message Joined: 28 Oct 04 Posts: 64 Credit: 34,444,555 RAC: 0 |
Update: Thursday the 12th Some progress has been made. Two systems still offline, five systems up, five systems still dropping. But it's still changing in smaller numbers than before which indicates that the scoring fix is not in yet. But now I am optimistic that progress is being made. Soon, I hope, the scoring fix will get us out of this strange situation. I suspect more problems than just scoring popped up, or they decided to rearrange/redesign some of the systems. Clearly the problem(s) were more complex than we suspected. BillN |
Send message Joined: 3 Nov 10 Posts: 39 Credit: 2,494,427 RAC: 0 |
from Bill "I suspect more problems than just scoring popped up, or they decided to rearrange/redesign some of the systems. Clearly the problem(s) were more complex than we suspected." as a retired IT data cruncher, i can testify that most ANY problem in IT turns out to be worse than you think, going in... IT rule(s) of thumb: a one-hour fix will take at least a day, by the time you figure out what is REALLY wrong... a one-day fix will probably take a week... a one-week fix will take a month, by the time you fix all the things you broke while you were fixing the original problem... a ninety-day conversion will take about a year, because the fool that made the first estimate doesn't know anything about the old system, the new system, or IT in general... been there - both as a witness and as a worker bee... |
Send message Joined: 31 Aug 04 Posts: 391 Credit: 219,896,461 RAC: 649 |
The latest two completed hadcm3n posted to my account went from 11,819.52 to 12,441.60 -- that means not a lot -- maybe an adjustment pending . Credits have not been evaluated nor posted to the credit sites for a few days. Like I say -- for the last few years working on this project -- they get it right in the end -- Right now with the "no wu available" if you have a fast machine try by prefs to get new (failed restarting) hadcmn3n -- they are out there patience required. As a 5 year cruncher here I expect that the new crew will figure how to work around the BOINC limitations , the hardware grant limitations, the disk space limitations, the OUC infrastructure limitations, the demands of the research grantees, and do a damn fine job of juggling the whole works. More open communication from the grantees and the infrastructure people would be welcome, but what can you say to so diverse an audience of contributors -- something I hope. Keep on crunching. E |
©2024 cpdn.org